oom-imagejob-fix

Signed-off-by: ashnamehrotra <[email protected]>
ashnamehrotra · Aug 28, 2023 · 5e63de1 · 5e63de1
1 parent f5942b1
commit 5e63de1
Show file tree

Hide file tree

Showing 11 changed files with 38 additions and 20 deletions.
diff --git a/.github/workflows/build-id.yaml b/.github/workflows/build-id.yaml
@@ -19,7 +19,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604 # v2.5.0
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09 # v2.5.1
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/codeql.yaml b/.github/workflows/codeql.yaml
@@ -24,7 +24,7 @@ jobs:
 
     steps:
     - name: Harden Runner
-      uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+      uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
       with:
         egress-policy: audit
 

diff --git a/.github/workflows/dep-review.yaml b/.github/workflows/dep-review.yaml
@@ -9,7 +9,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/deploy_docs.yaml b/.github/workflows/deploy_docs.yaml
@@ -1,4 +1,4 @@
-name: Deploy to GitHub Pages
+name: Generate docs website to GitHub Pages
 
 on:
   push:
@@ -7,13 +7,19 @@ on:
     paths:
       - '.github/workflows/deploy_docs.yaml'
       - 'docs/**'
+  pull_request:
+    branches:
+      - main
+    paths:
+      - '.github/workflows/deploy_docs.yaml'
+      - 'docs/**'
 
 permissions:
   contents: read
 
 jobs:
   deploy:
-    name: Deploy to GitHub Pages
+    name: Generate docs website to GitHub Pages
     runs-on: ubuntu-latest
     permissions:
       contents: write
@@ -24,7 +30,7 @@ jobs:
       - uses: actions/checkout@c85c95e3d7251135ab7dc9ce3241c5835cc595a9 # v3.5.3
 
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
 
@@ -49,6 +55,7 @@ jobs:
       - run: yarn build
 
       - name: Deploy to GitHub Pages
+        if: github.ref == 'refs/heads/main' && github.event_name == 'push' && github.repository == 'eraser-dev/eraser'
         uses: peaceiris/actions-gh-pages@373f7f263a76c20808c831209c920827a82a2847 # v3.9.3
         if: ${{ github.ref == 'refs/heads/main' }}
         with:

diff --git a/.github/workflows/e2e-build.yaml b/.github/workflows/e2e-build.yaml
@@ -14,7 +14,7 @@ jobs:
     timeout-minutes: 10
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - name: Set up Go 1.20
@@ -55,7 +55,7 @@ jobs:
     timeout-minutes: 10
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - name: Set up Go 1.20
@@ -95,7 +95,7 @@ jobs:
     timeout-minutes: 10
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - name: Set up Go 1.20
@@ -135,7 +135,7 @@ jobs:
     timeout-minutes: 10
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - name: Set up Go 1.20

diff --git a/.github/workflows/e2e-test.yaml b/.github/workflows/e2e-test.yaml
@@ -19,7 +19,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604 # v2.5.0
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09 # v2.5.1
         with:
           egress-policy: audit
 
@@ -47,7 +47,7 @@ jobs:
         E2E_TEST: ${{ fromJson(needs.build-e2e-test-list.outputs.e2e-tests) }}
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - name: Check out code into the Go module directory

diff --git a/.github/workflows/release-pr.yaml b/.github/workflows/release-pr.yaml
@@ -18,7 +18,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604 # v2.5.0
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09 # v2.5.1
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/release.yaml b/.github/workflows/release.yaml
@@ -20,7 +20,7 @@ jobs:
     timeout-minutes: 60
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/scorecard.yml b/.github/workflows/scorecard.yml
@@ -25,7 +25,7 @@ jobs:
 
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604 # v2.5.0
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09 # v2.5.1
         with:
           egress-policy: audit
 

diff --git a/.github/workflows/test.yaml b/.github/workflows/test.yaml
@@ -45,7 +45,7 @@ jobs:
     timeout-minutes: 40
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - uses: actions/checkout@c85c95e3d7251135ab7dc9ce3241c5835cc595a9 # v3.5.3
@@ -86,7 +86,7 @@ jobs:
     timeout-minutes: 40
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - name: Set up Go 1.20
@@ -118,7 +118,7 @@ jobs:
     timeout-minutes: 10
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
       - name: Check out code into the Go module directory
@@ -142,7 +142,7 @@ jobs:
       contents: read
     steps:
       - name: Harden Runner
-        uses: step-security/harden-runner@cba0d00b1fc9a034e1e642ea0f1103c282990604
+        uses: step-security/harden-runner@8ca2b8b2ece13480cda6dacd3511b49857a23c09
         with:
           egress-policy: audit
 

diff --git a/controllers/imagejob/imagejob_controller.go b/controllers/imagejob/imagejob_controller.go
@@ -419,6 +419,7 @@ func (r *Reconciler) handleNewJob(ctx context.Context, imageJob *eraserv1.ImageJ
 		if err != nil {
 			return err
 		}
+
 		log.Info("Started "+containerName+" pod on node", "nodeName", nodeName)
 		namespacedNames = append(namespacedNames, types.NamespacedName{Name: pod.Name, Namespace: pod.Namespace})
 	}
@@ -455,12 +456,22 @@ func (r *Reconciler) SetupWithManager(mgr ctrl.Manager) error {
 func podsComplete(podList []corev1.Pod) bool {
 	for i := range podList {
 		if podList[i].Status.Phase == corev1.PodRunning || podList[i].Status.Phase == corev1.PodPending {
-			return false
+			return containersFailed(&podList[i])
 		}
 	}
 	return true
 }
 
+func containersFailed(pod *corev1.Pod) bool {
+	statuses := pod.Status.ContainerStatuses
+	for i := range statuses {
+		if statuses[i].State.Terminated != nil && statuses[i].State.Terminated.ExitCode != 0 {
+			return true
+		}
+	}
+	return false
+}
+
 func (r *Reconciler) updateJobStatus(ctx context.Context, imageJob *eraserv1.ImageJob) error {
 	if imageJob.Name != "" {
 		if err := r.Status().Update(ctx, imageJob); err != nil {